Nebius Stock: An NVIDIA Badge, a Rival's Anthropic Win, and the Earnings Test Wall Street Keeps Deferring

Nebius won NVIDIA Exemplar Cloud status and bought Inferize, but a rival's Anthropic deal and OpenAI's revenue shortfall keep the return question open.

Two developments landed on Nebius this month that pull in opposite directions — one a technical credential, the other a reminder that demand in the AI infrastructure business is not distributed by default. Together they frame the question investors should be asking: not whether Nebius can build, but whether the build converts into earnings.

A rival draws first blood

Media reports say IREN signed a multi-year cloud computing agreement with Anthropic, a deal that fed directly into competitive anxiety around Nebius. The broader tech sector's weakness and lingering doubts about AI infrastructure spending did the rest, dragging the stock down 1.9% on the day.

That split matters. A soft tech tape says almost nothing about how well Nebius exploits its own capabilities. A competitor's signed contract, by contrast, speaks to who is actually capturing demand. Dismissing the pressure as mere market mood is the easy way out — and the wrong one.

Still, the IREN deal proves neither a lost order for Nebius nor a deterioration in its contract terms. It is, first and foremost, a win for a competitor. Reading it as a verdict on Nebius's entire business would be as hasty as ignoring the competitive question altogether. What counts for shareholders is not whether rival headlines weigh on the share price, but how Nebius answers them with its own offerings.

Inferize is the more concrete reply

The acquisition of Inferize, announced October 1, offers exactly that kind of answer. According to the company, Nebius folded the inference-optimization specialist's technology and team into its Nebius Token Factory. Transaction terms were not disclosed.

This is a play on the technical strength of the offering rather than a promise about the size of the AI market. It brings technology and personnel in-house — a concrete step, not a letter of intent. Even so, the deal should not be overread. No specific revenue contribution follows from it, and no demonstrated edge over IREN either. With financial terms under wraps, there is no way to judge what Nebius paid. The strategic direction is more convincing than any economic effect claimed today.

Sentiment got a boost from another quarter: analyst John McPeake of Rosenblatt Securities initiated coverage on October 8 with a Buy rating and a $304 price target. That is a positive assessment, not a proven business result — and after a competitor's win, a price target should not stand in for scrutiny of operating progress.

The demand side sends its own signal

A warning has surfaced on the customer side of the AI market. Reuters reported on October 8 that OpenAI told investors its annualized September revenue was nearly $50 billion, against roughly $70 billion previously held out. Those are not Nebius's numbers, and the shortfall justifies no immediate revision to its business expectations. It does, however, undercut the assumption that every growth projection in the AI space will be met on schedule.

The same Reuters report weighed on Nebius by sharpening concerns about returns on AI spending. No corresponding revenue gap at Nebius can be inferred — and that distinction is the whole point. The news concerns OpenAI first; its relevance to Nebius lies in a more critical appraisal of the entire investment landscape. It argues against the notion that every capacity-expansion milestone automatically justifies a higher valuation.

Technical wins, no blank check

Nebius did report a genuine technical advance on October 8: NVIDIA "Exemplar Cloud" status on NVIDIA HGX B300 for training workloads. That recognition concerns the technical side of the offering, and it matters because Nebius can now point to more than an expansion ambition. Investors should not stretch its meaning, though — a technical status answers a different question than the one about the business's economic viability. It says nothing about future margins or the profitability of added capacity.

The Rotterdam angle belongs in the same category. On Monday, Nebius signed a letter of intent with VOLT covering the lease of a substantial share of the planned capacity at the 600-megawatt Rotterdam AI gigafactory. No final capacity agreement was reached, and terms were not disclosed.

That puts Rotterdam on the opportunity list, not in the column of secured economic contributions. The project remains planned. Treating potential capacity as a committed build skips over a crucial distinction — one that should carry more weight in valuations precisely after the OpenAI news.

What the scorecard actually shows

Nebius deserves credit for tangible progress: the NVIDIA status and the Inferize acquisition argue for taking its operational development seriously. The non-binding Rotterdam arrangement, by contrast, calls for restraint on expectations that have run ahead of the paperwork.

The stance that follows is constructive on the technology and demanding on the economics. Nebius has earned recognition for concrete steps. A blanket vote of confidence for every further expansion plan is not warranted while the return debate has been reignited. What should decide the case is economic execution — not the scale of announced possibilities.